A problem space is the space of situations an Agent can be in and move between as it pursues its goals – the states it distinguishes, the transitions available to it, and the obstacles between where it is and where it is heading.
It is the world as the agent’s model carves it up, not the world in itself: an agent navigates the problem space its self-model constitutes, not a pre-given landscape (in Free Energy Minimisation terms, the model defines the envelope of viable states).
It is one of the three parts of the coherent self an Agent comprises, alongside a Cognitive Light Cone (the size of the goals it can pursue) and the cognitive processes – Intelligence – that navigate it.
Problem spaces are
- scale-relative. Each agent has its own problem space, and a higher-order agent’s differs from those of its parts: a movement’s strategic terrain is no single member’s. For an agent, its problem space is effectively its Environment as it can perceive and act on it.
- Complex Systems. The situations an agent faces are richly interconnected; mapping one bottom-up (as in Why-How Laddering or Concept Mapping) yields a Heuristic Device, never a complete description.
Human deliberation is a special case. The design-thinking distinction between the problem space (understanding the problem) and the solution space (generating answers) is what a problem space looks like at the scale of explicit human reasoning, as when a team frames ‘How Might We?’ questions.
References
- Newell & Simon (1972), Human Problem Solving (the origin of “problem space” in cognitive science)
- Levin (2019), “The Computational Boundary of a ‘Self’: Developmental Bioelectricity Drives Multicellularity and Scale-Free Cognition” (problem space as one element of the coherent self)